Estimating and testing rational expectations models when the trend specification is uncertain. (Q5958097)

This paper explores various strategies for estimating and testing rational expectations models when the trend specification is uncertain. One approach seeks to make estimators and tests robust to trend misspecification by reducing the influence of low frequency dynamics. However, contrary to intuition, the effects of trend specification errors are not confined to low frequencies, but are spread across the entire frequency domain. Thus, operations that damp low frequency components do not remove trend specification errors and are not sufficient for constructing robust estimators and tests. Another approach seeks representations of approximating models that do not condition on a specification of the trend, and it uses GMM to estimate parameters and test over-identifying restrictions. Because these methods do not condition on assumptions about trends, they are robust to errors in that part of the approximating model.